Neural Auto-Curricula
This work addresses the challenge of automating algorithm design in multi-agent reinforcement learning, offering a novel approach that could reduce human effort and improve performance in game-solving tasks, though it is incremental in automating existing principles.
The paper tackles the problem of manually designing update rules in multi-agent reinforcement learning for two-player zero-sum games by introducing Neural Auto-Curricula (NAC), a framework that uses meta-gradient descent to automate the discovery of learning rules, achieving competitive or better performance than state-of-the-art methods like PSRO across various games, including generalizing from small to large games such as from Kuhn Poker to Leduc Poker.
When solving two-player zero-sum games, multi-agent reinforcement learning (MARL) algorithms often create populations of agents where, at each iteration, a new agent is discovered as the best response to a mixture over the opponent population. Within such a process, the update rules of "who to compete with" (i.e., the opponent mixture) and "how to beat them" (i.e., finding best responses) are underpinned by manually developed game theoretical principles such as fictitious play and Double Oracle. In this paper, we introduce a novel framework -- Neural Auto-Curricula (NAC) -- that leverages meta-gradient descent to automate the discovery of the learning update rule without explicit human design. Specifically, we parameterise the opponent selection module by neural networks and the best-response module by optimisation subroutines, and update their parameters solely via interaction with the game engine, where both players aim to minimise their exploitability. Surprisingly, even without human design, the discovered MARL algorithms achieve competitive or even better performance with the state-of-the-art population-based game solvers (e.g., PSRO) on Games of Skill, differentiable Lotto, non-transitive Mixture Games, Iterated Matching Pennies, and Kuhn Poker. Additionally, we show that NAC is able to generalise from small games to large games, for example training on Kuhn Poker and outperforming PSRO on Leduc Poker. Our work inspires a promising future direction to discover general MARL algorithms solely from data.